【作者】 柳圣军；

【导师】 汤一平；

【作者基本信息】 浙江工业大学 ， 计算机应用技术， 2009， 硕士

【摘要】 水路智能交通系统是未来内河航运发展的必然趋势,对提高运输效率和保障运输安全有重要意义。计算机视觉技术的发展为水路交通系统提供了新的解决方法,利用视觉传感器可以获取范围广、信息量大的视频流,通过动态图像理解技术对视频流进行分析,可以获取当前交通实态信息。本文利用多传感器信息融合技术实现对内河航道交通流信息的获取。对近水域与远水域的广角摄像装置进行研究,利用广角摄像装置实现对内河航道的宏观监控,通过多目标跟踪算法实现对航道内的船舶实时跟踪,然后根据预先所定制的映射关系指示高速球摄像机对船名牌与船身进行微观抓拍并保存特写图,通过对船舶特写图的识别,解析出船舶船舷浮在水面的高度和船名牌号。接着以船名牌号为关键词检索船舶数据库得到该船舶的基本信息,最后利用船舶的基本信息以及解析出该船体的吃水深度估算该船舶的实际载重吨位。本文主要围绕水路交通智能监控系统的数字图像处理、模式识别和计算机视觉等关键技术,从低层视觉、中层视觉和高层视觉三个层次,对水路交通智能监控系统进行具体的研究。在低层视觉中,主要研究船舶的检测与分割算法,包括对图像的预处理、背景建模与更新、运动目标检测和连通区域标识等技术的研究;在中层视觉中,主要研究多目标船舶跟踪算法,以及交通流参数的检测;在高层视觉中,对船舶的实际载重吨位进行研究,并对船舶的异常行为预警进行探讨。本文开发了基于Java语言的内河航道智能监控系统,详细介绍各个模块的实现步骤,并在不同的环境下进行实验。实验证明该系统实时性强,准确度高并具有较好的鲁棒性。更多还原

【Abstract】 Waterborne Intelligent Transportation System(WITS) is an inevitable trend in inland river shipping in the future. It has an important significance to raise transportation efficiency and protect transportation safety. The development of Computer vision technology provides new solution for WITS: use vision sensor to get video flow with wide range and enormous information, then use dynamic image understanding technology to analyze the video flow, could get the real time circumstance information of the traffic.In this paper, we use multi-sensor information fusion technology to get the information of traffic flow in inland waterway, do research on the wide-angle Camera device of nearby waterway and faraway waterway, use wide-angle Camera device to deal with monitoring the inland waterway, and use multi-target tracking algorithm to track the ships in the waterway in real time. And then we capture the license plates and the hull of the tracked ships with the help of the high Speed Dome Camera according to the mapping relation customized previously. The captured ship images are recognized to determine license plate number and the height of shipboard which is higher than the horizontal plane. Then use the license plate number to retrieve the basic data from database. At last, we use the basic data and the draught of the ship to compute the real loading tonnage of the ship.In this paper, we mainly discuss the key techniques such as digital image processing, pattern recognition and computer vision for WITS. The specific study on WITS is discussed from three levels: low level vision, middle level vision and high level vision. The chief work in low level vision is ship detection and segmenting, including techniques of image preprocessing, background modeling and updating, moving object detection and connected region labeling, the chief work in middle level vision includes multi-target tracking algorithm and traffic flow parameters detection, and the chief work in high level vision includes computation of the real loading tonnage of the ship and abnormal behavior detection.This paper develops a inland waterway intelligent Surveillance system based on the Java language and introduces the steps of modules in detail and experiments in different environment. The results showed that the system has strong real-time, high accuracy and fairly good robustness.更多还原